IO concepts

The terms on this page form a conceptual model. They make architecture and operations discussions precise without claiming that IO implements a named API, protocol, or data model.

Execution vocabulary

Term	Meaning
Request	An authenticated proposal to run identified automation content with bounded inputs and targets
Authorization	The decision that permits a request under an approved policy and change context
Content identity	An immutable release, commit, digest, and source that identifies what may run
Input set	Non-secret parameters plus references to secrets resolved through an approved channel
Target scope	The explicit systems or abstract target class that a run may affect
Runtime identity	The workload identity and privileges used during execution
Run state	The runtime's observation of progress or termination; it is not target acceptance
Outcome verification	An independent observation of the intended target or service state
Run evidence	The classified, minimal record needed for traceability and review

Three separate decisions

Keep these decisions distinct:

1. May this request run? Authorization evaluates identity, policy, content, scope, timing, and risk.
2. Did execution complete? Runtime state records what happened inside the execution boundary.
3. Was the outcome accepted? Verification compares observed target state to the approved objective.

Conflating these decisions can turn a transport or process result into a false claim about service health.

Reproducibility and repeatability

A request is reproducible when its content, dependency, input-contract, and runtime identities can be resolved later. A run is repeatable only when the external conditions that affect it are also controlled. Neither property means an operation is safe to repeat after partial failure.

Evidence classification

Run records can include hostnames, identities, source paths, variable values, and failure output. A public status vocabulary may describe them, but real records remain in the authorized evidence system. Retain only what the environment's policy requires and protect it according to its classification.

See Architecture for the conceptual data flow and Security for trust-boundary decisions.